A coverage read you can defend
No black box. A channel either can or cannot carry a message to a person: a live token, a valid consent, a pulse of presence. Four plain-English eligibility rules, then every read is weighted by the value at the other end.
Who counts as reachable
Every customer who has purchased at least once. Registration captures an address; a first purchase proves it is real and read. The widest channel you own.
Reads: Registration record + first transaction
The app-first crowd: anyone using two or more products, anyone on complementary range, and newer cohorts who onboarded through the app with permissions still fresh.
Reads: Push token + breadth and tenure signals
The strongest relationships: subscription holders, plus frequent, higher-value customers. The people engaged enough to have trusted you with a mobile number.
Reads: Mobile number + subscription permissions
Anyone present in the product itself: complementary range users and broad customers. No consent wall and no inbox to fight through, but it only fires when they show up.
Reads: Session events + product breadth
Illustrative eligibility, inferred from behaviour so the demo lights up without new authoring. The production model reads your actual consent flags, contact tokens and deliverability history, channel by channel, Customer by customer.
Why every read is value-weighted
Headcount coverage lies by omission. A channel that reaches 70% of customers but misses your top tier is quietly failing where it costs the most. So every coverage number here carries the decayed 12-month spend of the people behind it, the same recency-weighted value the tiers use. Reachable audience tells you how many; value-weighted coverage tells you how much, and the gap between the two is the case for growing consent among the customers who matter.
No new tracking to invent. These are events you already capture; the model reads them as they are.
- ·Consent and preference records per channel, with source and timestamp
- ·Valid contact tokens: email address, push token, mobile number
- ·App instals, uninstals and session events
- ·Subscription contact permissions
- ·Preference-centre events: visits, topic choices, opt-downs
- ·Deliverability signals: bounces, complaints, dead addresses
All first-party and privacy-safe: consent records, contact tokens and presence signals you already hold. Nothing here depends on third-party cookies or bought data.
Models are tuned, not installed. Each step is earned on evidence, and the previous step keeps running until the next one beats it.
- v0Heuristic, day oneTransparent rules on the data you already have. Live in Braze in weeks, and everyone can see why it fired.
- v1Calibrated to your historyThresholds and weights re-fit to your own base and vertical, backtested against what actually happened.
- v2ML where it earns itA trained model replaces the rules only when it beats them on holdout data. Explainability stays a requirement.
- v3Monitored and re-tunedDrift watched, thresholds reviewed on a cadence, and every change proven with holdouts before it ships.
Reachability starts from the consent flags you already hold, so day one needs nothing new. It tunes with deliverability and sunset policies, so a dead address stops counting as reach. Then it grows into consent-capture experiments proven with holdouts, like the SMS activation play, so every point of opt-in is bought on evidence rather than hope.